Dear List members,
I am trying to run a selection model in which selection occurs at several
levels.
Let, y be a binary dependent variable with outcomes y0 and y1; N
observations. Then,
a) y1 is again a binary variable with outcomes y10 and y11; n1 observations
(n1<N)
b) y11 is a multinomial with outcomes y110, y111 and y112; n2 observations
(n2<n1)
c) y111 and y112 are binaries with outcomes {y1110; y1111; n3} and {y1120;
y1121; n4} where n3+n4<n2
d) y1111 (n=n5<n3) and y1121 (n=n6<n4) are continuous and our main equation
to be estimated.
Does anyone help me how y1111 and y1121 can be estimated in Stata?
Is the following approach acceptable both from theoretical and computational
viewpoints?
1) run logit with y; generate selection term (lambda1)
2) run logit with y1 including lambda1 as one of the regressors; generate
lambda11
3) run MNL with y11 including lambda11 as one of the regressors; generate
lambda111 and lambda112
4a) run logit with y111 incluidng lambda111 as one of the regressors;
generate lambda1111
4b) run logit with y112 incluidng lambda112 as one of the regressors;
generate lambda1121
5a) run OLS with y1111 including lambda1111 as one of the regressors
5b) run OLS with y1121 including lambda 1121 as one of the regressors
I would appreciate receiving your feedback.
Thanks in advance.
Subhash.
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
